Undergraduate Math 011

a firsT yeaR coursE in geometrY

Evan Chen

April Fool’s Day 2019

 Contents

Preface 3
0.1. Synopsis 3
0.2. Goals of this book 3
0.3. Casualties in achieving these goals 3
0.4. Satirical comments 4
0.5. Stray unironic comment 4

Chapter I. Synthetic geometry 5

Preliminaries 5
§1. Angles 5
1.1. Definitions 5
1.2. Cyclic quadrilaterals 7
1.3. Applications 7
1.4. Phantom points 8
1.5. Similar triangles 8
Exercises 8
§2. Power of a point 9
2.1. The power of a point 9
2.2. Radical axis 10
2.3. Radical center 11
Exercises 11
§3. Homothety 11
3.1. Definition and properties 11
3.2. Application to the medial triangle 12
3.3. Application to the nine-point circle 12
3.4. An application to tangent circles 13
Exercises 13
§4. Trigonometry and lengths 14
4.1. Notation 14
4.2. The extended law of sines 14
4.3. The law of cosines 14
Exercises 15
§5. Ceva and Menelaus 15
5.1. Signed areas 15
5.2. Ceva’s theorem 16
5.3. Menelaus theorem 16
Exercises 17
Chapter II. Coordinate systems 18
§6. Complex numbers 18
1
 CONTENTS 2

6.1. The complex plane 18

6.2. First results 18
6.3. The unit circle, and triangle centers 19
6.4. Triangle centers 19
Exercises 20
§7. Barycentric coordinates 21
7.1. Normalized coordinates 21
7.2. Coordinates of triangle centers 22
7.3. Homogenized coordinates 22
7.4. The vector plane and dot product 22
7.5. Displacement vectors 23
7.6. The equation of a circle 24
Exercises 24

Chapter III. Advanced techniques 26

§8. Inversion in the plane 26
8.1. Definition and first properties 26
8.2. Generalized lines and circles 26
8.3. Inversion distance formula 27
Exercises 28
§9. Projective geometry 29
9.1. Cross ratios 29
9.2. The case of cross ratio −1 30
9.3. An aside on projective transformations 30
9.4. Exercises 31
§10. Complete quadrilaterals 32
10.1. Spiral similarity 32
10.2. Miquel point of a cyclic quadrilateral 33
10.3. The Miquel point of a cyclic quadrilateral 34
Exercises 34

Index 35
 Preface

0.1. Synopsis. Among the four olympiad math subjects, geometry has the
reputation for being by far the most reliant on specific knowledge to do well in
(which haters of geometry often cite when complaining about it). But it occurred
to me recently that, despite this reputation, there isn’t really that much material
in the subject.
In the spring of 2016, I published a textbook Euclidean Geometry in Math
Olympiads (henceforth EGMO) which I think contains every theorem that an IMO
gold medalist would be “expected to know”. This has as far as I can tell become the
American standard for learning the subject, and these notes are based on EGMO.
I want to point out now — EGMO is pretty thin as far as things that could be
called a “textbook” go. The main matter is about 200 pages. And even then, most
of it is not theory — it is a lot of contest problems, worked examples, motivational
discussion, and so on.
So it occurred to me one day that all of olympiad geometry probably fits
comfortably in a typical one-semester college course. In fact, maybe not
even that. Half a semester might be more accurate, now that I think about it.
And so to stake my claim, I present to you my latest work:
Undergraduate Math 011: a firsT semesteR coursE in geometrY
Not offered this academic year.
0.2. Goals of this book. The intention of these notes is that they are
• self-contained,
• as short as possible, yet
• technically complete for olympiad purposes.
In particular, most major theorems in EGMO should appear, and the proofs (though
terse) should also be included for completeness when it is not unreasonable. It is
thus expected that any classical geometry problem appearing at IMO or USAMO
will fall entirely in the span of the techniques contained in this document (and per-
haps even within the first chapter). The lecture notes are divided into 10 sections,
which loosely correspond in order to the 10 chapters of EGMO.
0.3. Casualties in achieving these goals. In achieving these goals, while
making the text as short as possible, the following decisions were made:
• No friendly speech or discussion. We are in the fast lane, and everything
is correspondingly terse.
• No “worked examples” (okay, there are a few, but not many). The only
material presented is the core material; most “applications” are exercises.
• No diagrams. This is a little extreme, but I do remember once trying to
read a graph theory textbook that had no figures… so not unheard of.
3
 PREFACE 4

• Few named theorems. No named problems. (I believe this is the norm.)

• Few contest problems. Most exercises are chosen to be boring “textbook-
style” ones.1
• In particular, important theorem are often left as exercises.2 In some cases
things which would be theorems in EGMO are downgraded to exercises
here, to maximize confusion with inessential results.
• No solutions. We are a math textbook here: “left as an exercise”.
0.4. Satirical comments. As a consequence of these design decisions, olympiad
contestants will to some extent find these notes useless, and should read EGMO in-
stead. Indeed part of my point in writing this was to draw attention to the contrast
between
(i) having a surface-level understanding of an area of math, and
(ii) being able to solve IMO-level problems with it.
In some ways, this is a celebration of just how strong the top high school
geometers are. More pessimistically, it can also be considered a tongue-in-cheek
jab at how higher math is often taught. For this reason, I also took pride in
making it look like the type of writing style and formatting that is common in
real mathematics — that is, the usual stuffy, formal, bleed-your-eyes-in-boredom
writing that passes for exposition in mediocre textbooks.
In short, here is what olympiad geometry might look like if it was taught in a
typical top-tier university.
0.5. Stray unironic comment. I considered not including this preface in
order to make this document more of an April Fool’s Day gag. But then I realized
that there was an uncomfortably real possibility that people might not realize this
was a satirical document and try to use it, and that would be very bad.

1Incidentally, there is something qualitatively unappealing about the 101 exercises I chose for
these lecture notes. See, a geometry problem which appears on an olympiad seems to me to have
a personality: it is a concrete, specific problem, that is unlikely to be independently constructed
by someone else. Whereas the exercises in these lecture notes are nameless, unmemorable, and
somehow all look the same.
2For the record: the more I think about it, the more I object to this common practice in
certain textbooks. I increasingly suspect it is just due to laziness by authors. The “main matter”
is (in my opinion) not the right time to let the reader “work it out for themselves”, precisely
because they are seeing material for the first time. They should eventually be able to work out
some proofs for themselves — but not now.
 CHAPTER I

Synthetic geometry

Preliminaries
We will generally assume a knowledge of very basic high school geometry, for
example the axioms of Euclid.
Let ABC be a triangle. We also take for granted the existence of four remark-
able triangle centers:
(1) The incenter I is the intersection of the internal angle bisectors, which is
also the center of the unique circle inside the triangle tangent to all three
sides.
(2) The centroid G is the unique point inside the triangle such that ray AG
bisects side BC, etc.
(3) The orthocenter H is the intersection of the three altitudes of the triangle.
(4) The circumcenter O is the center of the unique circle passing through A,
B, and C.
Proofs of the existence of these centers can be found in any standard reference,
and we will also supply proofs in the exercises. In general, if we state a result
involving a triangle ABC then we may refer to any of the names G, H, I, O
without unnecessary redefinition.
We also give names to the associated inscribed triangles.
(1) The intouch triangle of ABC is the triangle whose vertices are the contact
points of the incircle with the sides of ABC.
(2) The medial triangle of ABC is the triangle whose vertices are the mid-
points of the sides of ABC.
(3) The orthic triangle of ABC is the triangle whose vertices are the feet of
the altitudes.
We also remind the reader that by (XY Z) we mean the circle through X, Y ,
Z.

§1. Angles
1.1. Definitions. For most of this book, we are going to be working with
so-called directed angles.
1.1.1. Definition. Given any two non-parallel lines ` and m, we define the
directed angle
](`, m)
to be the measure of the angle starting from ` and ending at m, measured coun-
terclockwise. If ` and m are parallel or coincide, we define the angle to be zero.
Moreover, unless specified otherwise, the measures of the angles are always consid-
ered modulo 180◦ .
5
 §1. ANGLES 6

From here, it is easy to check the following results.

1.1.2. Proposition. Let `, m, n be lines.
(1) ](`, m) = −](m, `).
(2) If `, m, n are concurrent, then ](`, m) + ](m, n) = ](`, n).
(3) If m k n and ` is another line, then ](`, m) = ](`, n).
We leave the proof as Exercise 3.
The angle formed by three points then follows as a special case of this.
1.1.3. Definition. Given three points A, O, B we define


]AOB := AO, BO .
Equivalently, if ` and m are two lines which intersect at O, then ](`, m) = ]AOB
for any point A on ` and B on m.
One can then translate the earlier proposition to obtain several results in the
new notation, which will be used more frequently.
1.1.4. Proposition. For any distinct points A, B, C, P in the plane, we have
the following rules.
(1) ]AP A = 0.
(2) ]ABC = −]CBA.
(3) ]AP B + ]BP C = ]AP C.
(4) ]P BA = ]P BC if and only if A, B, C are collinear.
(5) If AP ⊥ BP , then ]AP B = ]BP A = 90◦ .
(6) ]ABC + ]BCA + ]CAB = 0.
(7) AB = AC if and only if ]ACB = ]CBA.
(8) If AB k CD, then ]ABC + ]BCD = 0.
Proof. (1) ]AP A = ](AP , AP ) = 0.
(2) ]ABC = ](AB, BC) = −](CB, BA) = −]CBA.
(3) ]AP B + ]BP C = ](AP , P B) + ](BP , P C) = ](AP , CP ) = ]AP C.
(4) By using (2) and (3)
together, we have
]P BA − ]P
BC = ]P BA +
]CBP = ] CB, BP + ] BP , ]BA = ] CB, BA . This is equal to
zero if and only if lines CB and BA coincide, which is equivalent to A,
B, C collinear.
(5) This is obvious.
(6) Introduce a line ` through A parallel to BC (this is possible by Euclid’s
fifth postulate). Then
]ABC + ]BCA + ]CAB




= ] AB, BC + ] BC, CA + ] CA, AB




= ] AB, ` + ] `, CA + ] CA, AB



= ] AB, CA + ] CA, AB
= 0.
(7) This is equivalent to the fact that an triangle is isosceles if and only if two
of the vertex angles are equal.
(8) ](AB, BC) + ](BC, CD) = ] AB, CD = 0.



 §1. ANGLES 7

1.2. Cyclic quadrilaterals.

1.2.1. Definition. We say that a set of points is concyclic if one can draw a
circle through them. Four concyclic points are said to form a cyclic quadrilateral.
Since any three non-collinear points are concyclic (as we will prove later on),
this definition is only interesting with four or more points.
We state the following important theorem (which will be invoked repeatedly).
1.2.2. Theorem. Four distinct points A, B, X, Y are concyclic if and only if
]AXB = ]AY B.
Regrettably, its proof is case-dependent and hence rather tiresome, so we omit
the proof from this book. However, the utility of this theorem cannot be overstated.
In a degenerate case when two points coincide, we also have the following
equivalence.
1.2.3. Theorem. Line ` passes through a point P on a circle centered at O.
The circle also passes through A and B. Then the following are equivalent:
(1) ` is tangent to the circle.
(2) ` ⊥ OP .

(3) ] `, P A = ]P BA.
1.3. Applications. We give some first examples of applications of our results
now, which can be applied ubiquitously.
1.3.1. Lemma. Let ABC be a triangle with orthic triangle DEF . Then the
quadrilaterals AEHF , BF HD, CDHE, BF EC, CF DA, ADEB are concyclic.
Proof. The quadrilateral AEHF is cyclic since ]AEH = ]AEB = 90◦ =
]AF C = ]AF H. The cases BF HD and CDHE are similar.
The quadrilateral BF ED is cyclic since ]BF C = 90◦ = ]BEC, and similarly
for the next three. 
1.3.2. Lemma. Let ABC be a triangle and D, E, F points on lines BC, CA,
AB. Then (AEF ), (BF D), (CDE) are concurrent at a point.
Proof. Let K be the second intersection of the two circles. Then ]F KD =
]F BD = ]ABC and ]DKE = ]DCE = ]BCA. As 0 = ]DKE + ]EKF +
]F KD = ]ABC + ]BCA + ]CAB, it follows ]EKF = ]CAB, and so AEKF
is concyclic. 
1.3.3. Definition. The point of concurrence in the previous lemma is referred
to as the Miquel point of DEF with respect to ABC.
That being said, because we are taking modulo 180◦ , we most be careful to
not take “half” of a directed angle, since in that situation things may not make
sense. Here is an example of a situation in which it is thus necessary to pay some
attention to the positions of the points.
1.3.4. Definition. We define the A-excenter of triangle ABC as the inter-
section of the ∠A bisector and the external angle bisectors of ∠B and ∠C. It is
typically denoted Ia .
1.3.5. Lemma. Let ABC be a triangle.
 §1. ANGLES 8

(i) The midpoint of IIa is the arc midpoint of arc BC

d not containing A.
(ii) The points B, I, C, Ia lie on a circle with diameter IIa .
Since this is routine, the proof is left as an exercise.
1.4. Phantom points. Moving forward, it will often be helpful to be able to
rephrase problems in certain ways. Although this does not need official definition,
we shall give an example of a situation in which it might be helpful.
1.4.1. Lemma. Let ABC be a triangle with AB 6= AC. The perpendicular
bisector of BC intersects the ∠A bisector at a point K. Then the points A, B, C,
K are concyclic.
Proof. It turns out that there is not an easy way to directly show the angles
with the point K was given. However, suppose we define K 0 as the intersection of
the ∠A bisector with the circumcircle. Then (say by Lemma 1.3.5) it follows that
K 0 is the midpoint of arc BC,
d so BK 0 = CK 0 , and K 0 lies on the perpendicular
bisector of BC too. This implies K = K 0 , since two lines can only intersect at a
unique point. 
1.5. Similar triangles. We mention now, as we will have need for it later,
that similar triangles can also be written in terms of directed angles. In doing so
we may also pay attention to their orientations.
1.5.1. Definition. Consider triangles ABC and XY Z. We say they are directly
similar if
]ABC = ]XY Z, ]BCA = ]Y ZX, and ]CAB = ]ZXY.
We say they are oppositely similar if
]ABC = −]XY Z, ]BCA = −]Y ZX, and ]CAB = −]ZXY.
If either condition is satisfied we say they are similar.
The upshot of this is that we may continue to use directed angles when proving
triangles are similar; we just need to be a little more careful. In any case, as you
probably already know, similar triangles also produce ratios of lengths.
1.5.2. Proposition. The following are equivalent for triangles ABC and XY Z.
(i) 4ABC ∼ 4XY Z.
(ii) (AA) ∠A = ∠X and ∠B = ∠Y .
(iii) (SAS) ∠B = ∠Y , and AB : XY = BC : Y Z.
(iv) (SSS) AB : XY = BC : Y Z = CA : ZX.
Thus, lengths (particularly their ratios) can induce similar triangles and vice
versa. However, notice that SAS similarity does not have a directed form; see the
exercises.
Exercises.
1. Exercise. If A, B, C are points lying on a circle centered at O, show that
]AOC = 2]ABC.
2. Exercise. If A, B, C, D are four distinct points, show that
]ABC + ]BCD + ]CDA + ]DAB = 0.
 §2. POWER OF A POINT 9

3. Exercise. Finish the proof of Proposition 1.1.2.

4. Exercise. Find an example of two triangles ABC and XY Z such that
AB : XY = BC : Y Z, ]BCA = ]Y ZX, but 4ABC and 4XY Z are not similar.
5. Exercise. Identify the Miquel point associated to (a) the orthic triangle,
(b) the intouch triangle.
6. Exercise. Show that triangle Ia Ib Ic has orthic triangle ABC.
7. Exercise (*). Prove Lemma 1.3.5. Why does it require undirected angles?
8. Exercise. Let ABCDE be a convex pentagon such that BCDE is a square
with center O and ∠A = 90◦ . Show that AO bisects ∠BAE.
9. Exercise. Let P be a point inside circle ω. Consider the set of chords of ω
that contain P . Prove that their midpoints all lie on a circle.
10. Exercise. In cyclic quadrilateral ABCD, let I1 and I2 denote the incenters
of 4ABC and 4DBC, respectively. Show that quadrilateral I1 I2 BC is cyclic.
11. Exercise. Let ABC be an acute triangle inscribed in circle Ω. Let X be
the midpoint of the arc BC
d not containing A and define Y , Z similarly. Identify
the orthocenter of 4XY Z.
12. Exercise. Let ABC be a triangle with altitudes BE, CF and let M denote
the midpoint of BC. Show that the tangents to (AEF ) at E and F meet at M ,
and the tangent to (AEF ) at A is parallel to BC.
13. Exercise (*). Let ABC be a triangle and P a point on its circumcircle.
Show that feet of the perpendiculars from P to sides BC, CA, AB are collinear.
This line is called the Simson line of P .
14. Exercise (*). Let ABC be a triangle with intouch triangle DEF . Let M
and N denote the midpoints of BC and AC, and let K denote the intersection of
BI and EF . Show that ∠BKC = 90◦ and moreover K lies on line M N .

§2. Power of a point

In addition to behaving well with respect to angles, circles give rise to length
relations which give a powerful tool for our study in geometry.
2.1. The power of a point. The basis of the power of a point theorem starts
in the following observation.
2.1.1. Theorem. Let P be a point inside a circle. Two lines through P are
drawn. The first line intersects the circle at A and A0 (which may be the same
point if the line is a tangent). The second line intersects the circle at B and B 0
(which may be the same point if the line is a tangent). Then
P A × P A0 = P B × P B 0 .
Proof. This is an easy application of similar triangles. Observe that ]P B 0 A =
]BB 0 A = ]BA0 A = ]BA0 P . Similarly, ]P A0 B = ]AB 0 P . Therefore triangles
P A0 B and P B 0 A are similar and oppositely oriented. Consequently, PPAB = PPBA
0 0

which rearranges to the given result. 

In light of this, it is convenient to make the following definition.
 §2. POWER OF A POINT 10

2.1.2. Definition. Let ω be a circle with radius r and center O. The power
of P with respect to ω, which we denote as Powω (P ), is defined as P O2 − r2 .
Note that:
• If P lies on ω, then Powω (P ) = 0.
• If P lies outside ω, and A, A0 are as in the previous theorem, then
Powω (P ) = P A × P A0 . This follows by taking the corresponding B
and B 0 so that BB 0 is a diameter of the circle, whence P B × P B 0 =
(P O + r)(P O − r) = P O2 − r2 .
• If P lies inside ω, and A, A0 are as in the previous theorem, then Powω (P ) =
−AP × P A0 , by a similar argument.
Therefore this definition gives an “intrinsic” way of defining the fixed quantity in
Proposition 2.1.1.
There is another way to realize the sign convention above.
2.1.3. Definition. Let A, P , A0 be collinear points. We agree that the signed
product P A × P A0 has negative sign if P lies between A and A0 , and positive
otherwise. We also adopt the convention AP = −P A, so that both AP × P A0 =
−P A × P A0
2.1.4. Definition. The signed quotient PA
P A0 A0 follows the same sign
= − PAP
convention.
Then all three cases above can be succinctly summarized:
2.1.5. Theorem. If a line through a point P intersects a circle ω at two points
A and A0 , then
P A × P A0 = Powω (P ).
Note that there is no longer a distinction between whether P lies inside or
outside the circle.
With the signed product convention, the power of a point theorem has a con-
verse too.
2.1.6. Theorem. Two lines AA0 and BB 0 meet at a point P such that
P A × P A0 = P B × P B 0 .
Then A, A0 , B, B 0 are cyclic.
We leave the proof as Exercise 15.
2.2. Radical axis. Suppose now we consider two circles.
2.2.1. Theorem. Let ω and γ be two non-cocentric circles. Then the locus
of points P with Powω (P ) = Powγ (P ) is a line perpendicular to the line through
their centers.
Proof. We have a rare application of Cartesian coordinates. Let us denote
by (a, 0) the center of the first circle, with radius r. Also let us denote by (b, 0) the
center of the second circle, with radius s. Then
Powω (P ) = Powγ (P )
⇐⇒ [(x − a) + y 2 ] − r2 = [(x − b)2 + y 2 ] − s2
2

⇐⇒ 2(a − b)x = b2 − a2 + r2 − s2
which describes a vertical line. 
 §3. HOMOTHETY 11

Note that if the two circles have common points, those common points have
power 0 and are thus on the locus. In particular, if two circles meet at points A
and B, it is more economical to describe their radical axis as line AB.

2.3. Radical center. We are given now three circles ω1 , ω2 , ω3 whose centers
are not collinear. We can draw the radical axis of any pair.
2.3.1. Theorem. In this situation the three radical axes are concurrent.
Proof. Suppose the radical axis of ω1 and ω2 meets the radical axis of ω2 and
ω3 at P . (Because the centers are not collinear, the radical axes are not parallel.)
Then Powω1 (P ) = Powω2 (P ) = Powω3 (P ). So P lies on the third radical axis. 

2.3.2. Definition. The concurrency point is called the radical center of the
three circles.
This gives a valuable way to show that lines are concurrent which often cannot
easily be done with just angle chasing alone

Exercises.
15. Exercise. Prove Theorem 2.1.6.
16. Exercise. Let t > 0. If Powω (P ) = t2 > 0 give a geometric interpretation
for t.
17. Exercise. Three circles are drawn, each pair having a common chord.
Show that their common chords are concurrent.
18. Exercise (*). Let ABC be a triangle with circumradius R and inradius
r. Show that IO2 = R(R − 2r).
19. Exercise. Two circles meet at A and B. A line is drawn tangent to the
first circle at X, and tangent to the circle at Y . Show that line AB bisects XY .
20. Exercise. Let A, B, C be three noncollinear points, and draw a circle of
radius zero at each. Determine their radical center. Use this to give a proof that
the circumcenter O exists.
21. Exercise. Identify the radical axis of the circles with diameter AB and
AC. Use this to prove that the orthocenter H exists.
22. Exercise (*). Scalene triangle ABC is given with orthic triangle DEF .
Show that circles (AOD), (BOE), (COF ) meet at a point other than O.
23. Exercise (**). Let ABCD be a quadrilateral and let E = AB ∩ CD,
F = BC ∩ DA. Show that the circles with diameters AC, BD and EF have a
common radical axis. Deduce that the midpoints of AC, BD, EF are collinear.
(Hint: consider the orthocenters of certain triangles.)

§3. Homothety
3.1. Definition and properties. One way to capture at once a lot of in-
formation that normally would form a similar triangles argument is through the
notion of homothety.
 §3. HOMOTHETY 12

3.1.1. Definition. A homothety h is a transformation defined by a center O

and a nonzero real number k (not necessarily positive). It sends a point P to
another point h(P ), multiplying the distance from O by k.
It is important to note that k can be negative; in that case, O will lie between
P and h(P ).
It is easy to see that homothety preserves similarity. Homothety also preserves
many things, including but not limited to tangency, angles (both vanilla and di-
rected), circles, and so on. They do not preserve length, but they work well enough:
the lengths are simply all multiplied by k.
3.1.2. Proposition. Given noncongruent parallel segments AB and XY , there
is a unique homothety sending A to X and B to Y .
Proof. Since AB 6= XY , the quadrilateral ABY X is not a parallelogram, and
we may take O to be the intersection of lines AX and BY . Then 4OAB ∼ 4OXY .
The common scale factor is then the signed quotient OXOA = OB .
OY

This is often used with triangles. A consequence of this is the following useful
lemma.
3.1.3. Corollary. Let ABC and XY Z be non-congruent triangles such that
AB k XY , BC k Y Z, and CA k ZX. Then lines AX, BY , CZ concur at some
point O, and O is a center of a homothety mapping 4ABC to 4XY Z.
Proof. Exercise 24. 
3.2. Application to the medial triangle. Suppose that 4DEF is the me-
dial triangle of 4ABC, then the corresponding sides are parallel and have ratio 12 .
Consequently:
3.2.1. Theorem. The centroid G of a triangle is the center of a negative ho-
mothety, with ratio − 12 , which maps ABC to its medial triangle.
This immediately gives applications.
3.2.2. Corollary. If D is the midpoint of BC, then we have the signed quo-
tient GD
GA
= −2.
3.2.3. Corollary. Points G, H, O are collinear with signed quotient GH
GO =
−2.
Proof. Observe that since DO ⊥ BC, we have DO ⊥ EF . Therefore O
is actually the orthocenter of 4DEF , so the homothety will map it to H, the
orthocenter of 4ABC. 
3.3. Application to the nine-point circle. Here is yet another application.
First, we need the following fact.
3.3.1. Proposition. In triangle ABC,
(a) the reflection of H across side BC lies on (ABC).
(b) the reflection of H across the midpoint of side BC lies on (ABC) diamet-
rically opposite A.

Proof. This is just angle chasing, see Exercise 25. 

 §3. HOMOTHETY 13

3.3.2. Theorem. Let ABC be a triangle. There is a circle passing through the
following nine points:
• The midpoints of AH, BH, CH.
• The midpoints of BC, CA, AB.
• The feet of the altitudes.
The center of this circle is the midpoint of OH, and this circle has half the radius
of (ABC).
Proof. Consider a homothety at H with ratio 2. Each of the nine points
mentioned goes to a point on (ABC), the first three by definition, the latter six
by Proposition 3.3.1. Thus after the homothety, the nine points lie on (ABC). So
before the homothety, they must have been on a circle as described. 
3.3.3. Definition. This circle is called the nine-point circle. Its center is called
the nine-point center.
3.3.4. Definition. The Euler line is the line through the points G, O, H and
the nine-point center.
3.4. An application to tangent circles. Tangents at circles are also nicely
handled by homothety.
3.4.1. Proposition. Chord AB is given in a circle Ω. Let ω be a circle tangent
to chord AB at K and internally tangent to ω at T . Then ray T K passes through
the midpoint M of arc AB
d of Ω, not containing T .
Proof. Since Ω and ω are tangent at T , it follows there is a homothety at T
taking ω to Ω. Because the tangent to ω at K (which is line AB) is parallel to the
tangent to Ω at M it follows the homothety maps K to M . Hence T , K, M are
collinear. 
Exercises.
24. Exercise. Prove Corollary 3.1.3.
25. Exercise. Prove Proposition 3.3.1.
26. Exercise. Given two noncongruent circles, neither contained in the other,
show that there is a unique positive homothety sending one to the other, and its
center is the intersection of the common external tangents.
27. Exercise. Let ABC be a triangle. The incircle touches BC at D, while
the A-excircle touches BC at E. Show that AE passes through the antipode of D
on the incircle.
28. Exercise. In the notation of the previous exercise, show that EI bisects
the A-altitude.
29. Exercise (*). Consider three circles ω1 , ω2 , ω3 in the plane no two con-
gruent and with disjoint interiors. For each pair of circles, we construct the inter-
section of their common external tangents. Prove that these three intersections are
collinear.
30. Exercise (*). In triangle ABC with contact triangle DEF , point M is
the midpoint of BC. Prove that the lines AM , EF , DI are concurrent. (Hint:
draw a line through DI ∩ EF and let it meet sides AB and AC at B 0 and C 0 . Try
to show AB 0 IC 0 is cyclic and B 0 I = C 0 I.)
 §4. TRIGONOMETRY AND LENGTHS 14

31. Exercise (**). Let ABC be a triangle and D be a point on AB. Suppose
a circle ω is tangent to CD at L, AB at K, and also to (ABC). Show that the
incenter of ABC lies on line LK.

§4. Trigonometry and lengths

We assume the reader is familiar with the notation sin and cos. These are tools
which allow one to relate sides and angles together, and accordingly become quite
useful. Since these topics are regularly covered in high school curriculums, we will
be brief.

4.1. Notation. Throughout this section (and moving forwards), if 4ABC is

a given triangle, we will let a = BC, b = CA, c = AB denote the lengths of the
triangle, and A = ∠BAC, B = ∠ABC, C = ∠BCA denote the measures of the
angles. It is also customary to set
a+b+c
s=
2
as the so-called semiperimeter of the triangle.

4.2. The extended law of sines. The law of sines is usually given in a
sin A =
. . . but there is in fact an extended form which is more symmetric.

4.2.1. Theorem. If R denotes the circumradius of 4ABC, then

a b c
2R = = = .
sin A sin B sin C
A proof, as well as several applications, is given in the exercises. Note that this
means that if we know two angles of a triangle and one side, then we can recover
everything else.

4.3. The law of cosines. We would like to also be able to determine a triangle
based on two side lengths and the included angle, and the law of cosines does this
for us.

4.3.1. Theorem. We have

c2 = a2 + b2 − 2ab cos C.

Proof. We handle only the case where cos C > 0 since the other cases are
analogous. The A-altitude to side BC, which we denote by AD, has AD = b sin C
and CD = b cos C. Then by the Pythagorean theorem,

c2 = AD2 + BD2
= (b sin C)2 + (a − b cos C)2
= b2 (sin C)2 + a2 − 2ab cos C + b2 (cos C)2
= a2 + b2 (sin C)2 + (cos C)2 − 2ab cos C
 

= a2 + b2 − 2ab cos C. 
 §5. CEVA AND MENELAUS 15

Exercises.

32. Exercise. Deduce the Pythagorean theorem from the law of cosines.

33. Exercise. The ∠A bisector meets side BC at D. Show that DB/DC =

AB/AC.

34. Exercise. Show that a = 2R sin A in Theorem 4.2.1, thus establishing the
theorem.

35. Exercise. Let ABC be a triangle and D a point on side BC. Show that

a × (DA2 + DB × DC) = b2 × DB + c2 × DC.

36. Exercise. Let A1 A2 A3 A4 A5 be a convex pentagon. Assume rays Ai+1 Ai+2

and Ai+4 Ai+3 meet at point Xi . Show that
5
Y 5
Y
Xi Ai+2 = Xi Ai+3 .
i=1 i=1

Here all indices are taken modulo 5.

37. Exercise (*). In a convex cyclic quadrilateral ABCD show that

AB · CD + BC · DA = AC · BD.

38. Exercise (*). In this problem we derive a formula for the area of a triangle
in terms of a, b, c.
(a) Show that the area of ABC is given by 12 ab sin C.
(b) Express sin C as a function of a, b, c.
(c) Prove that the area of ABC is given by
p
s(s − a)(s − b)(s − c).
√
39. Exercise (*). Show that the area of a triangle is given also by rra rb rc ,
where r is the radius of the incircle and ra , rb , rc are the radii of the excircles.

§5. Ceva and Menelaus

5.1. Signed areas. We saw earlier that we got some convenience from signed
quotients of lengths, and we will in what follows want to adopt a similar convention
for signed areas. Thus we make the following definition.

5.1.1. Definition. If P , Q, R are three points then we denote by [P QR] the

signed area of triangle P QR; its absolute value is the same as the usual area, but
• it is zero if P , Q, R are collinear,
• it is positive if P , Q, R are situated in counterclockwise order on its
circumcircle, and
• it is positive if P , Q, R are situated in clockwise order on its circumcircle.

Note that unlike with signed quotients and products, we can specify a sign of
a single area (rather than, say, a quotient of areas).
 §5. CEVA AND MENELAUS 16

5.2. Ceva’s theorem. Ceva’s theorem is a useful result which allows one to
determine concurrence of so-called cevians of a triangle: lines joining the vertex of
a triangle to a point on the opposite side. The proof will use our convention of
directed quotients.
5.2.1. Theorem. Let ABC be a triangle and D, E, F points on the opposite
sides or their extensions, which are distinct from the vertices. Then lines AD, BE,
CF are concurrent if and only if
BD CE AF
× × = 1.
DC EA F B
Proof. First let us assume that the three cevians are indeed concurrent at P .
Then we have
BD [P BD] [ABD]
= =
DC [P CD] [ACD]
where the areas are signed. This means that
BD [ABD] − [P BD] [ABP ]
= = .
DC [ACD] − [P CD] [ACP ]
Multiplying cyclically yields the desired equation.
For the converse, assume that the desired concurrence holds. We let lines BE
and CF meet at P , and let line AP meet side BC at D0 . Then BD D 0 C = DC , which
0
BD

is enough to imply D = D . 0


For example, this immediately implies that the centroid of the triangle exists,
because the signed ratios above are all equal to 1. We leave the remaining centers
as a good exercise.
There is also a “trigonometric” form of the same theorem, in case it is easier
to access the angles. For simplicity we only state it with the cevians inside the
triangle.
5.2.2. Theorem. Let ABC be a triangle and D, E, F points on the opposite
sides. Then lines AD, BE, CF are concurrent if and only if
sin ∠BAD sin ∠CBE sin ∠ACF
× × = 1.
sin ∠CAD sin ∠ABE sin ∠BCF
Proof. We have by the law of sines that
AD
sin ∠BAD BD × sin B BD sin C
= AD
= ×
sin ∠CAD DC × sin C
DC sin B
and when we multiply this cyclically, we find that this reduces to Ceva’s theorem.


5.3. Menelaus theorem. There is a similar theorem if we are hoping that

D, E, F are collinear (rather than lines AD, BE, CF being concurrent).
5.3.1. Theorem. Let ABC be a triangle and D, E, F points on the opposite
sides or their extensions, which are distinct from the vertices. Then points D, E,
F are concurrent if and only if
BD CE AF
× × = −1.
DC EA F B
 §5. CEVA AND MENELAUS 17

Proof. As with Ceva’s theorem, there is no hope of the collinearity if the

sign is positive, and hence we can focus on the case of negative sign. Moreover,
by a similar argument with a point D0 , it suffices to prove the forwards direction,
assuming D, E, F are collinear.
Let Z = AD ∩ BE and define X, Y analogously. Applying Ceva’s theorem on
4ACD (with cevians meeting at F ) and 4ABD (with cevians meeting at E) that
AY DB CE
1= × ×
YD BC EA
AF BC DZ
1= × × .
FB CD ZA
If we write down the four analogous expressions and multiply them all together, we
will find that  2
BD CE AF
1= × ×
DC EA F B
which, since we are only considering the case of negative sign, implies the result. 
Exercises.
40. Exercise. Use Ceva’s theorem to establish the existence of the incenter I.
41. Exercise. Use Ceva’s theorem to establish the existence of the orthocenter
H.
42. Exercise (*). Let AD, BE, CF be cevians meeting in the interior of
4ABC. Let DX, EY , F Z be cevians meeting in the interior of 4DEF . Show
that lines AX, BY , CZ are concurrent.
43. Exercise. Let AD, BE, CF be cevians meeting at a point P in the
interior of 4ABC. Denote by D0 the reflection of D across the midpoint of BC
and define E 0 , F 0 similarly. Show that cevians AD0 , BE 0 , CF 0 are concurrent.
(The concurrency point is called the isotomic conjugate of P .)
44. Exercise (*). Let P be a point in the interior of 4ABC. Show that there
exists a point Q inside 4ABC such that ]BAP = ]QAC, ]CBP = ]QBA,
]ACP = ]QCB. (The point Q is called the isogonal conjugate of P .)
45. Exercise. What is the isogonal conjugate of the orthocenter?
46. Exercise (*). Let ABC be a triangle and let s = 12 (a + b + c) denote its
semiperimeter. We also let D, E, F be the contact points of the incircle on the
opposite sides.
(a) Prove that BD = s − b and CD = s − c.
(b) Show that lines AD, BE, CF are concurrent. The concurrency point is
called the Gergonne point.
47. Exercise (**). In triangle ABC the A-excircle touches side BC at D, and
points E, F are defined similarly. Show that lines AD, BE, CF are concurrent.
The concurrency point is called the Nagel point.
 CHAPTER II

Coordinate systems

§6. Complex numbers

6.1. The complex plane. Let C and R denote the set of complex and real
numbers, respectively. Our intention is to assign a complex number z ∈ C to each
point Z in the plane.
Each z ∈ C can be expressed as
z = a + bi = r (cos θ + i sin θ) = reiθ
√
where a, b, r, θ ∈ R and 0 ≤ θ < 2π. We write |z| = r = a2 + b2 and arg z = θ.
Recall that
|z1 z2 | = |z1 | |z2 | and arg z1 z2 = arg z1 + arg z2 .
More importantly, each z is associated with a complex conjugate z = a −
bi, which commutes with addition and multiplication We also have the following
important identity.
6.1.1. Proposition. For any complex number z, |z| = zz.
2

Proof. This is obvious by expanding. 

When doing geometry problems, it is customary that the coordinates of a point
denoted by a capital letter are represented by the lowercase letter; for example the
complex coordinate of A is represented a. We adopt this convention without further
comment in what follows.
6.2. First results.
6.2.1. Proposition. Let A, B, C, D be pairwise distinct points. Then AB ⊥
CD if and only if b−a
d−c
∈ iR; i.e.
 
d−c d−c
+ = 0.
b−a b−a
 
Proof. It’s equivalent to d−c
b−a ∈ iR ⇐⇒ arg d−c
b−a ≡ ±90◦ ⇐⇒ AB ⊥
CD. 
6.2.2. Proposition. Let A, B, C be pairwise distinct points. Then A, B, C
are collinear if and only if c−a
c−b ∈ R; i.e.
 
c−a c−a
= .
c−b c−b

Proof. Similar to the previous one. 

18
 §6. COMPLEX NUMBERS 19

6.2.3. Proposition. Let A, B, C, D be pairwise distinct points. Then A, B,

C, D are concyclic if and only if
c−a d−a
: ∈ R.
c−b d−b
   
Proof. It’s not hard to see that arg c−a
c−b = ∠ACB and arg d−a
d−b = ∠ADB.
(Here angles are directed). 
6.3. The unit circle, and triangle centers. On the complex plane, the unit
circle is of critical importance. Indeed if |z| = 1 we have
1
z= .
z
This means that we can get the following useful formulas.
6.3.1. Lemma. If A and B lie on the unit circle, then then the foot from Z to
AB is
1
(z + a + b − abz) .
2
Proof. By earlier propositions, we want a point w such that
w−z w−z
=−
a−b 1/a − 1/b
w−a w−a
=
w−b w−b
and solving the system for w gives the result above. 
6.3.2. Lemma. If A, B, C, D lie on the unit circle then the intersection of AB
and CD is given by
ab(c + d) − cd(a + b)
.
ab − cd
Proof. By earlier propositions, we want a point w such that
w−a w−a
=
w−b w−b
w−c w−c
=
w−d w−d
and again solving gives the result. 
These are much easier to work with than the corresponding formulas in general.
6.4. Triangle centers. We can also obtain the triangle centers immediately:
6.4.1. Theorem. Assume points A, B, C lie on the unit circle. Then the
circumcenter, centroid, and orthocenter of ABC are given by 0, 13 (a+b+c), a+b+c,
respectively.
Observe that the Euler line follows from this.
Proof. The results for the circumcenter and centroid are immediate. Let
h = a + b + c. By symmetry it suffices to prove AH ⊥ BC. We may set
h−a b+c
z= = .
b−c b−c
 §6. COMPLEX NUMBERS 20

Then   1 1
b+c b+c b + c c+b
z= = = 1 1 = = −z
b−c b−c b − c
c−b
so z ∈ iR as desired. 
There is also a result for the incenter. It is more complicated and we will not
use it, so we do not prove it here.
6.4.2. Theorem. Points A, B, C lie on the unit circle. Then one can assign
a choice of complex numbers x, y, z on the unit circle such that a = x2 , b = y 2 ,
c = z 2 and
(a) the arc midpoints opposite the vertices of the triangle are −yz, −zx, −xy,
and
(b) the incenter has coordinate −(xy + yz + zx).
Exercises.
48. Exercise. Let W be the reflection of Z across AB. Show that
(a − b)z + ab − ab
w= .
a−b
49. Exercise (*). Show that the signed area of 4ABC is given by the deter-
minant  
 a a 1 
i 
b b 1  .
4
c c 1 
50. Exercise. Find the coordinates of the nine-point center of a triangle ABC
inscribed in the unit circle, in terms of a, b, c.
51. Exercise. Points A and B lie on the unit circle which are not diametrically
opposite. Show that the tangents at A and B meet at a point with coordinates
a+b .
2ab

52. Exercise. Show that if A and B lie on the unit circle, then a point P lies
on line AB if and only if
p + abp = a + b.
53. Exercise. In Theorem 6.4.2, show that (b) follows from (a).
54. Exercise (*). Let ABCD be a cyclic quadrilateral. Let HA , HB , HC ,
HD denote the orthocenters of BCD, CDA, DAB, ABC. Show that AHA , BHB ,
CHC , DHD are concurrent.
55. Exercise (**). Let ABC be a triangle and P a point on its circumcircle.
Show that the Simson line of P bisects P H.
56. Exercise. Let A, B, C, D be points. Show that lines AB and CD intersect
at
(āb − ab̄)(c − d) − (a − b)(c̄d − cd)¯
¯ .
(ā − b̄)(c − d) − (a − b)(c̄ − d)
57. Exercise (*). Let ABC be a triangle and erect equilateral triangles
on sides BC, CA, AB outside of ABC with centers OA , OB , OC . Prove that
4OA OB OC is equilateral and that its center coincides with the centroid of triangle
ABC.
 §7. BARYCENTRIC COORDINATES 21

58. Exercise (**). Prove that the Euler lines of triangles IAB, IBC, ICA,
ABC are concurrent.

§7. Barycentric coordinates

We present a second coordinate system. In this chapter, 4ABC is a fixed non-
degenerate reference triangle with vertices in counterclockwise order. The lengths
will be abbreviated a = BC, b = CA, c = AB. These correspond with points in
the vector plane A,
~ B,
~ C.
~

7.1. Normalized coordinates. Each point in the plane is assigned an or-

dered triple of real numbers P = (x, y, z) such that

P~ = xA
~ + yB
~ + zC
~ and x + y + z = 1.

These are called the barycentric coordinates of the point.

The most important result about these coordinates is that one can recover
signed areas using them. (For this reason, they are sometimes called areal coor-
dinates instead.) We state this in the following theorem. To avoid technicalities
about how to define the notion of “area” rigorously, we do not include the proof.

7.1.1. Theorem. Let P = (xP , yP , zP ), Q = (xQ , yQ , zQ ), R = (xR , yR , zR ).

The signed area of a triangle [P QR] is given by
 
 xP yP zP 
 
[P QR] =  xQ yQ zQ .

 xR yR zR 

This theorem right away gives an equivalent, useful definition for the coordi-
nates of point P .

7.1.2. Corollary. For each point P ,

 
[BP A] [CP B] [AP C]
P = , , .
[ABC] [ABC] [ABC]
Another important corollary is that equations of a line take the form of linear
equations.

7.1.3. Theorem. Let u, v, w be real numbers not all equal. Then the locus of
points (x, y, z) satisfying
ux + vy + wz = 0
is a straight line, and moreover all lines are of this form.

Proof. One direction follows by fixing two points Q and R in Theorem 7.1.1
noting that the locus of points lying on line QR is precisely those points with
[P QR] = 0.
Conversely, suppose WLOG that u 6= v and u 6= w. Then one may take
−v
Q = ( u−v u
, u−v , 0) and R = ( u−w
−w u
, 0, u−w ) and line QR will have the desired form
in Theorem 7.1.1. 
 §7. BARYCENTRIC COORDINATES 22

7.2. Coordinates of triangle centers. We now work out the coordinates of

some common triangle centers. The remaining ones can be found in the exercises.
7.2.1. Proposition. The barycentric coordinates of the centroid are G =
(1/3, 1/3, 1/3).
Proof. The centroid is given by 13 (A
~+B
~ + C).
~ 
7.2.2. Proposition. The barycentric coordinates of the incenter are I =
a b c
( a+b+c , a+b+c , a+b+c ).
First proof. Note that [BIC] = ar, where r is the inradius, whilst [CIA] =
br and [AIB] = cr. So it follows from Corollary 7.1.2. 
Second proof. By the angle bisector, point I is collinear with A = (1, 0, 0)
and the point D = (0, b+cb c
, b+c ).
The equation of line AD is then seen to be cy − bz = 0. Indeed, it fits the
format of Theorem 7.1.3 and it passes through both of the desired points.
So the second and third coordinate of I are in a b : c ratio. Proceeding in a
cyclic fashion gives the result. 
7.3. Homogenized coordinates. One may by now notice that the denomi-
nators are quite cumbersome. So, the following convention will be convenient.
7.3.1. Definition. Let u, v, w be real numbers
 with nonzero sum. Then
(u : v : w) is shorthand for u+v+w , u+v+w , u+v+w .
u v w

We distinguish these conventions with colons and commas respectively. The

former will be called homogeneous barycentric coordinates while the latter (original)
is normalized barycentric coordinates. For example, we may now conveniently write
I = (a : b : c), which is shorter and more informative.
7.3.2. Remark. It’s nice to note that, because in Theorem 7.1.3 the terms are
all degree one in x, y, z these homogeneous coordinates work out of the box, as
well.
As an example, let’s consider the incenter
 
a b c
I = (a : b : c) = , , .
a+b+c a+b+c a+b+c
Moreover let ` : ux + vy + wz = 0 be a line. If we want to see whether I lies on `,
the condition is
u·a+v·b+w·c=0
a b c
⇐⇒ u · +v· +w· = 0.
a+b+c a+b+c a+b+c
This is nearly the same as saying we can use the coordinates (a : b : c) directly in
calculations.
7.4. The vector plane and dot product. To make further progress we
adopt the following conventions. We have been working with vectors in the plane
of ABC; at this point, we agree that the circumcenter of the triangle ABC will be
designated as the origin This means that A,~ B,
~ C~ each have magnitude equal to
the circumradius R of 4ABC; in symbols,

~ ~ ~
A = B = C = R.
 §7. BARYCENTRIC COORDINATES 23

7.4.1. Definition. Recall that the dot product · is defined as follows: if ~v and
~ are vectors and the angle between them is θ then the dot product is defined as
w
the real number
~v · w
~ = k~v k kwk
~ cos θ.
The dot product is commutative and distributes over addition.
The dot product has the following properties.
7.4.2. Proposition. (i) We have ~v · ~v = k~v k .
2

(ii) We have ~v · w
~ = 0 if and only if the vectors are perpendicular.
We can also compute the dot products of our basis vectors.
7.4.3. Proposition. We have A ~ = R2 and A
~ ·A ~ = R2 − c2 /2, and the cyclic
~ ·B
variations.
Proof. The first part is clear. The second calculation can be realized as:
~·B
A ~ = R2 cos ∠AOB
= R2 cos 2C
= R2 1 − 2 sin2 C

1 2
= R2 − (2R sin C)
2
= R2 − c2 /2. 

Thus we can compute the dot products of any vectors expressed as sums of A,
~
~ ~
B, C.
7.5. Displacement vectors. We may now compute the distance between any
two points. For convenience, given two points P and Q we can consider their
displacement vector, thought of as the vector from P to Q.
7.5.1. Definition. A displacement vector of two (normalized) points P =
−−→
(p1 , p2 , p3 ) and Q = (q1 , q2 , q3 ) is denoted by P Q and is equal to (q1 − p1 , q2 −
p2 , q3 − p3 ).
Note that the sum of the coordinates of a displacement vector is 0.
7.5.2. Theorem. Let P and Q be two arbitrary points and consider a displace-
−−→
ment vector P Q = (x, y, z). Then the distance from P to Q is given by
2
|P Q| = −a2 yz − b2 zx − c2 xy.
Proof. Observe that
   
2 ~ + yB
~ + zC
~ · xA~ + yB
~ + zC
~ .
|P Q| = xA
Applying the properties of the dot product
2
X X
|P Q| = ~·A
x2 A ~+2 ~·B
xy A ~
cyc cyc
 
X 1
= R2 (x2 + y 2 + z 2 ) + 2 xy R2 − c2 .
cyc
2
 §7. BARYCENTRIC COORDINATES 24

Collecting the R2 terms,

2
|P Q| = R2 (x2 + y 2 + z 2 + 2xy + 2yz + 2zx) − (c2 xy + a2 yz + b2 zx)
= R2 (x + y + z)2 − a2 yz − b2 zx − c2 xy
= −a2 yz − b2 zx − c2 xy. 
7.6. The equation of a circle. Since a circle is the locus of points equidistant
from a point, we ought to be able to work out its equation. In fact, we can say
slightly more.
7.6.1. Theorem. Let ω be a circle. Then there exist real numbers u, v, w such
that for any point P = (x, y, z) we have
Powω (P ) = −a2 yz − b2 zx − c2 xy + (ux + vy + wz)(x + y + z).
In particular, the points of ω are those for which Powω (P ) = 0.
We choose to present the equation in this form so that the right-hand side is
entirely degree 2 as a polynomial in x, y, z, even though we have x + y + z = 1.
This is so that homogeneous coordinates can be used to determine whether a point
lies on the circle; the denominators factor immediately (cf. Remark 7.3.2).
Proof. Assume the circle has center O = (x0 , y 0 , z 0 ) and radius r. Then
Powω (P ) = P O2 − r2
= −a2 (y − y 0 )(z − z 0 ) − b2 (z − z 0 )(x − x0 )
− c2 (x − x0 )(y − y 0 ) − r2
= −a2 yz − b2 zx − c2 xy + λ1 x + λ2 y + λ3 z + λ4
where λ1 , λ2 , λ3 , λ4 are constants. We may then take u = λ1 + λ4 , v = λ2 + λ4 ,
w = λ3 + λ4 . 
As with Theorem 7.1.3, one can determine the equation of a circle by selecting
any particular points. See Exercise 61.
Exercises.
59. Exercise. Use barycentric coordinates to give alternate proofs of Ceva and
Menelaus.
60. Exercise (*). Let ABC be a triangle and let P be a point in its interior.
Lines P A, P B, P C intersect sides BC, CA, AB at D, E, F , respectively. Prove
that
1
[P AF ] + [P BD] + [P CE] = [ABC]
2
if and only if P lies on at least one of the medians of triangle ABC.
61. Exercise. Show that the equation of the circumcircle of ABC is given by
−a2 yz − b2 zx − c2 xy = 0.
62. Exercise. Find the coordinates of the A-excenter.
63. Exercise. Use Theorem 7.6.1 to give another proof that the radical axis
of two circles is a straight line.
64. Exercise. Show:
 §7. BARYCENTRIC COORDINATES 25

(a) The barycentric coordinates of the circumcenter are given by

O = (sin 2A : sin 2B : sin 2C) .
(b) The barycentric coordinates of the orthocenter are given by
H = (tan A : tan B : tan C) .
65. Exercise. Use the previous exercise to give another proof that the ortho-
center, circumcenter, and centroid are collinear.
66. Exercise (*). Find the equation of the tangent to the circumcircle of
triangle ABC at A.
67. Exercise. Find the coordinates of the Gergonne and Nagel points (defined
in Exercises 46 and 47). Show that the Nagel point is collinear with the incenter
and centroid.
68. Exercise (*). Prove that if P = (x : y : z) is a point inside triangle ABC,
then its isogonal conjugate (defined in Exercise 44) is given by P ∗ = (a2 /x : b2 /y :
c2 /z).
69. Exercise. Show that the perpendicular bisector of BC has equation
0 = a2 (z − y) + x(c2 − b2 ).
−−→ −−→
70. Exercise. Let M N = (x1 , y1 , z1 ) and P Q = (x2 , y2 , z2 ) be displacement
vectors. Show that M N ⊥ P Q if and only if
0 = a2 (z1 y2 + y1 z2 ) + b2 (x1 z2 + z1 x2 ) + c2 (y1 x2 + x1 y2 ).
71. Exercise (**). Let A1 A2 A3 A4 be a non-cyclic quadrilateral. For 1 ≤ i ≤
4, let Oi and ri be the circumcenter and the circumradius of triangle Ai+1 Ai+2 Ai+3
(where Ai+4 = Ai ). Prove that
1 1 1 1
+ + + = 0.
O1 A21 − r12 O2 A22 − r22 O3 A23 − r32 O4 A24 − r42
 CHAPTER III

Advanced techniques

§8. Inversion in the plane

In what follows, we consider the usual Euclidean plane R2 with an additional
“point at infinity”, which we denote as ∞. We consider every line to pass through
this point ∞.

8.1. Definition and first properties.

8.1.1. Definition. For a circle ω with center O and radius r > 0, an inversion
around ω is a map that sends each point P to be a point P ∗ as follows.
• If P = ∞, then P ∗ = O.
• If P = O, then P ∗ = ∞.
• For any other point P , we choose P ∗ to be the unique point satisfying
OP · OP ∗ = r2 .
We immediately identify some properties of the inversion.
8.1.2. Proposition. Inversion is an involution: (P ∗ )∗ = P .
8.1.3. Proposition. The point P lies on ω if and only if P ∗ lies on ω.
In the case where P lies outside the circle, there is also a geometric interpreta-
tion.
8.1.4. Theorem. Let P be a point outside ω and suppose P A, P B are tangents
to the circle. Then P ∗ coincides with the midpoint of AB.
The proof of this theorem is left as Exercise 72.

8.2. Generalized lines and circles. We continue to fix a circle ω with center
O and radius r, through which we will perform inversions.
If ` is a line, by its inverse `∗ we mean the set
`∗ = {P ∗ | P ∈ `} .
Similarly for a circle γ its inverse is the set
γ ∗ = {P ∗ | P ∈ γ} .
The main result of this section is that inverses of lines and circles are themselves
lines and circles.
We check this using the following propositions.
8.2.1. Proposition. If ` is a line passing through O, then `∗ = `.
Proof. This is immediate by definition. 
26
 §8. INVERSION IN THE PLANE 27

8.2.2. Proposition. If γ is a circle passing through O, its inverse γ ∗ is a line

not passing through O. Dually, if ` is a line not passing through O, its inverse `∗
is a circle passing through O.

Proof. By an appropriate homothety, we may reduce to the case where γ

intersects the circle ω at two points A and B, and thus r = OA = OB. Let C be
any point on γ now and let D = AB ∩ CO. Then we have ]ACO = ]ABO =
]OAB = ]OAD, and thus 4ACO ∼ 4OAD, so OC · OD = OA2 . Thus D = C ∗ ,
ergo C ∗ lies on line AB. In other words γ ∗ is contained in line AB. It is not hard
to see that every point on line AB can be obtained this way, and also γ ∗ should
contain ∞ since γ contained O, so we are done. 

8.2.3. Proposition. If γ is a circle not passing through O, its inverse γ ∗ is a

circle not passing through O.

Proof. Suppose a line from O meets γ at A and B, and a second line meets
O at C and D. Then letting t2 = OA · OB = OC · OD we quickly deduce rt2 =
4

OA∗ · OB ∗ = OC ∗ · OD∗ , and so by power of a point from O the four points A∗ ,

B ∗ , C ∗ , D∗ are cyclic as well. If we treat AB as fixed and vary C, D, (noting hat
t does not change either) we obtain the result. Note the circle γ ∗ does not contain
O since γ does not pass through ∞. 

This can be succinctly summarized as follows. We say a generalized circle is

either a circle or a line: thus we may think of a usual line as a “circle passing
through ∞”.

8.2.4. Theorem. The inverse of a generalized circle is itself a generalized circle.

Then, all the qualifiers about “passing through O or not” and “passing through
∞ or not” can be captured in this statement.

8.3. Inversion distance formula. We have the following additional result,

sometimes known as the “distance formula”.

8.3.1. Theorem. In the usual notation, we have

r2
A∗ B ∗ = · AB.
OA · OB
Here is one application, generalizing Exercise 37 from before.

8.3.2. Theorem. For any four points in the plane A, B, C, D we have

AB · CD + BC · DA ≥ AC · BD.

Equality holds if ABCD is cyclic, in that order.

Proof. Consider an inversion about the circle centered at D with radius r.

On the one hand we have by the triangle inequality

A∗ B ∗ + B ∗ C ∗ ≥ A∗ C ∗
 §8. INVERSION IN THE PLANE 28

and the inequality is sharp if ABCD was cyclic in that order. However, we have
that
r2
A∗ B ∗ = · AB
DA · DB
r2
B∗C ∗ = · BC
DB · DC
r2
A∗ C ∗ = · AC
DA · DC
which rearranges to the inequality. 

We give many more applications of inversion in the exercises. Typically, one

can get interesting results if one inverts around a point with many circles passing
through it, and carefully analyzing the results.

Exercises.
72. Exercise. Prove Theorem 8.1.4.
73. Exercise. Show that we always have ]OAB = −]OB ∗ A∗ .
74. Exercise (*). Let ABC be a triangle with intouch triangle DEF . Show
that the inverse of the circumcircle of 4ABC with respect to the incircle coincides
with the nine-point circle of 4DEF . Conclude that the Euler line of 4DEF passes
through O if 4ABC is not equilateral.
75. Exercise. Show that if 4ABC has circumcenter O and we invert around
the circle centered at C with radius r, then O∗ is the reflection of C across A∗ B ∗ .
76. Exercise. Give another proof of Theorem 8.1.4 by inversion around the
circle centered at M with radius M B = M C. Deduce additionally that M K ·M T =
M B2 = M C 2.
77. Exercise. We say two circles γ and ω are orthogonal if they intersect at
two points A and B, and the tangents to γ and ω at A are perpendicular. Show
that in that case, if we invert around ω, then γ ∗ = γ.
78. Exercise. Let ABC be a right triangle with ∠C = 90◦ and let X and Y be
points in the interiors of CA and CB, respectively. Construct four circles passing
through C, centered at A, B, X, Y . Prove that the four points lying on at exactly
two of these four circles are concyclic.
79. Exercise (*). Let ABCD be a quadrilateral whose diagonals are perpen-
dicular and meet at E. Prove that the reflections of E across the sides of ABCD
are concyclic.
80. Exercise (*). Let ω1 , ω2 , ω3 , ω4 be circles with consecutive pairs tangent
at A, B, C, D. Prove that quadrilateral ABCD is cyclic.
81. Exercise (**). Let ABC be triangle whose A-excircle touches BC at
E. Suppose γ is a circle tangent to AB, to AC, and internally tangent to the
circumcircle of 4ABC at T . Prove that ]BAT = ]EAC.
 §9. PROJECTIVE GEOMETRY 29

§9. Projective geometry

This time, we consider the usual Euclidean plane R2 with an additional “point
at infinity” along each set of parallel lines, rather than just one point at infinity.
Thus, any two lines intersect at a unique point; the point is a usual Euclidean
point if they are not parallel, but it is considered to intersect at an infinity point
otherwise.
9.1. Cross ratios. We begin with the following definition.
9.1.1. Definition. Given four distinct collinear points A, B, X, Y (which may
be at infinity) we define the cross ratio as:
XA YA
(AB; XY ) = ÷ .
XB YB
Note that the cross ratio has the following uniqueness property.
9.1.2. Proposition. Given four distinct collinear points A, B, X, Y (which
may be at infinity) and let Y 0 be a fifth point different from A, B, X. Assume that
(AB; XY ) = (AB; XY 0 ). Then Y = Y 0 .
It is possible also to define the cross ratio of four concurrent lines, in the
following way.
9.1.3. Definition. Suppose four distinct lines a, b, x, y are concurrent at some
point P . Their cross ratio is then defined as
sin ∠(x, a) sin ∠(y, a)
(ab; xy) = ÷ .
sin ∠(x, b) sin ∠(y, b)
Some care is required to describe the signs in the above definition; it is more
economical to declare that the sign of the entire cross ratio is positive if one of the
four angles formed by line a and b contains both x and y, and negative otherwise.
The key result is the following equivalence.
9.1.4. Theorem. Suppose four distinct lines a, b, x, y are concurrent at some
point P . A line ` distinct from P meets a, b, x, y at four points A, B, X, Y . Then
(AB; XY ) = (ab; xy).
This is shown using the law of sines, see Exercise 85. It has widespread appli-
cations. For example, the immediate corollary is itself very useful.
9.1.5. Corollary. Let ` and `0 be two lines and P a point not on either line.
Let A, B, X, Y be four points on `. Then let A0 be the intersection of line P A and
`0 and define B 0 , X 0 , Y 0 similarly. Then
(AB; XY ) = (A0 B 0 ; X 0 Y 0 ).

When applying this corollary it is customary to use the notation (AB; XY ) =

P

(A0 B 0 ; X 0 Y 0 ) to indicate the point through which we are projecting.

We now state without proof one more result that will allow us to analyze conics:
we define the cross ratio of four points relative to a circumconic of those four points.
9.1.6. Definition. Let γ be a nondegenerate conic. Then four points A, B,
X, Y on γ we define (AB; XY )γ = (P A, P B; P X, P Y ) where P is any fifth point
on the conic; this doesn’t depend on the choice of P .
 §9. PROJECTIVE GEOMETRY 30

Here is a nice application of the results so far, the so-called Pascal theorem.
9.1.7. Theorem. Let A, B, C, D, E, F be six points inscribed in a conic γ.
Define X = AB ∩ DE, Y = CD ∩ F A, Z = BC ∩ EF . Then X, Y , Z are collinear.
Proof. Let P = AB ∩ EF and Q = DE ∩ AF and let Z 0 = XY ∩ EF . Then
B D X
(EP ; ZF ) = (EA; CF )γ = (QA; Y F ) = (EZ 0 ; P F )
so Z 0 = Z. 
9.2. The case of cross ratio −1. The most important special case is the
situation of four points with cross ratio −1, which is sometimes known as a harmonic
cross ratio. To make progress on these we use state the following useful lemma,
whose proof is trivial, hence left as an exercise.
9.2.1. Lemma. If A, B, X, Y are distinct collinear points such that (AB; XY ) =
(AB; Y X) then (AB; XY ) = −1.
We now give two examples of situations in which the cross ratio of −1 occurs.
9.2.2. Proposition. Let ABC be a triangle with concurrent cevians AD, BE,
CF . Line EF meets BC at X. Then (XD; BC) = −1.

Proof. Let Y = AD ∩ EF . Then (BC; XD) = (F E; XY ) = (BC; DX).

A P

9.2.3. Proposition. Let γ be a nondegenerate conic and X and Y two points
on the conic. Suppose the tangents to γ at X and Y meet at P . Consider another
line through P meeting γ at two points A and B. Let Q = AB ∩ XY . Then
(a) (AB; P Q) = −1 and
(b) (AX; BY )γ = −1.

Proof. We have (AB; P Q) = (AB; XY ) = (AB; QP ).

X Y

One nice application of Lemma 9.2.3 is that we can make the following defini-
tion.
9.2.4. Definition. Let γ be a nondegenerate conic and P a point. Consider
lines through P intersecting γ at X, Y and let Q be the unique point on line XY
with (XY ; P Q) = −1. Then the locus of Q lies on a line called the polar of P (with
respect to γ). Point P is the pole of Q.
9.2.5. Remark. If P X, P Y are tangents from P to γ, then the polar of P
coincides with line XY .
Two more examples of this situation are given in Exercise 82 and 89.
9.3. An aside on projective transformations. This is a rich topic which is
largely beyond the scope of these notes, so we merely mention some famous results.
9.3.1. Definition. A projective collineation is a bijection of the projective
plane which preserves collinearity of points.
The following result is sometimes called the “fundamental theorem of projective
geometry”.
9.3.2. Theorem. Let τ be a projective collineation. Then the image of any
conic under τ is a conic. Moreover cross ratios are preserved in the following senses:
 §9. PROJECTIVE GEOMETRY 31

• If A, B, C, D are four collinear points and τ maps them to W , X, Y , Z

then (AB; CD) = (W X; Y Z).
• If a, b, c, d are four concurrent lines and τ maps them to w, x, y, Z then
(ab; cd) = (wx; yz).
• If A, B, C, D are four points on a conic γ and τ maps these points to W ,
X, Y , Z and the conic γ to γ 0 then (AB; CD)γ = (W X; Y Z)γ 0 .
The following existence theorem guarantees the existence of homographies sat-
isfying certain conditions.
9.3.3. Theorem. (a) There exists a unique homography taking any four
points, no three collinear, to any other quadruple of four points with no
three collinear.
(b) Let γ be a circle or ellipse and P a point inside it. One can find a unique
homography mapping every point of γ to a (possibly different) point of γ,
while mapping P to the center of γ.
This means many results which can be stated only in projective terms can be
proved in special cases by taking a suitable homography. For example, one can
prove Proposition 9.2.2 in the following way: consider a homography mapping A,
B, C, and BE ∩ CF to an equilateral triangle and its center. If we let B 0 , C 0 , D0 ,
X 0 be the images of B, C, D, X then we have (B 0 C 0 ; D0 X 0 ) = −1 since D0 is the
midpoint of B 0 C 0 while X 0 is the point at infinity along line B 0 C 0 . Therefore we
actually have (BC; DX) = (B 0 C 0 ; D0 X 0 ) = −1 as needed.
We won’t use this technique at all, but it is worth noting that such proofs are
possible.
9.4. Exercises.
82. Exercise. Let M be the midpoint of a segment AB and ∞ the point at
infinity along line AB. Show that (AB; M ∞) = −1.
83. Exercise. Verify Lemma 9.2.1.
84. Exercise. Show that if γ is a circle and convex, quadrilateral AXBY is
inscribed in it, then in fact
   
 XA   Y A 
(AB; XY )γ = − 
  ÷  .
XB   Y B 
85. Exercise. Prove Theorem 9.1.4 using the law of sines.
86. Exercise. Give another proof of Lemma 9.2.2 by using Ceva and Menelaus
theorems.
87. Exercise. Give another proof of Lemma 9.2.3 in the case where γ is a
circle by using similar triangles and Exercise 84.
88. Exercise (*). Let ABXC be a quadrilateral inscribed in circle ω with
(AX; BC)ω = −1. Let M be the midpoint of AC.
(a) Show that ]BAM = ]XAC.
(b) Prove that line AX is concurrent with the tangents to ω at B and C,
possibly at infinity.
(The line AX is a symmedian of 4ABC.)
 §10. COMPLETE QUADRILATERALS 32

89. Exercise (*). Let X, A, Y , B be collinear points in that order and let C
be any point not on this line. Show that any two of the following conditions implies
the third condition.
(i) (AB; XY ) − 1;
(ii) ∠XCY = 90◦ ; and
(iii) CY bisects ∠ACB.

90. Exercise. Prove that if γ is a circle with center O and P is a point outside
it, then the polar of P is the line through the inverse P ∗ which is perpendicular to
line OP .

91. Exercise. Let γ be a nondegenerate conic. Show that point P lies on the
polar of Q if and only if Q lies on the polar of P .

92. Exercise. Let AB, CD, P Q be chords of a conic γ concurrent at M . Let

X = P Q∩AD, Y = P Q∩BC, then prove that (P M ; XQ) = (P Y ; M Q). Conclude
that if M
MP
Q = −1 then M Y = −1.
MX

93. Exercise (**). Let ABCD be a quadrilateral inscribed in nondegenerate

conic γ. Set P = AB ∩ CD, Q = BC ∩ DA, and R = AC ∩ BD. Prove that P ,
Q, R are the poles of QR, RP , P Q, respectively. (Hint: use Proposition 9.2.2 and
Proposition 9.2.3 repeatedly.)

94. Exercise. Let ABCD be a quadrilateral and γ any circumconic. Let

XY ZW be a quadrilateral and ω any circumconic. Assuming that (AB; CD)γ =
(XY ; ZW )ω , there exists a (unique) homography sending ABCD to XY ZW and
which maps each point of γ to a point of ω.

95. Exercise (*). Let A, B, C, D be four points, no three collinear. Define

the points P = AD ∩ BC, Q = AB ∩ CD, and R = AC ∩ BD. Let X1 , X2 , Y1 , Y2
denote P R ∩ AD, P R ∩ BC, QR ∩ AB, QR ∩ CD. Prove that lines X1 Y1 , X2 Y2 ,
and P Q are concurrent.

§10. Complete quadrilaterals

If we say ABCDEF is a complete quadrilateral, we mean that A, B, C, D are
four points, no three collinear, and the points E = AB ∩ CD, F = BC ∩ DA exist.
The complete quadrilateral is cyclic if A, B, C, D lie on a circle; we may then refer
to the circumcircle and center of this circle.

10.1. Spiral similarity.

10.1.1. Definition. A spiral similarity with a center O combines a rotation

about O with a dilation.

The most commonly occurring case of a spiral similarity is between two seg-
ments, in which case we have direct similarity. We prove the relevant existence
result now:

10.1.2. Theorem. Let AB and CD be segments such that ABDC is not a

parallelogram. There exists a unique spiral similarity taking A to C and B to D.
 §10. COMPLETE QUADRILATERALS 33

Proof. We determine O in terms of A, B, C, D via complex numbers. The

similarity we want is equivalent to:
c−o d−o
= .
a−o b−o
Solving gives
ad − bc
o= .
a+d−b−c
The denominator is nonzero so we are done. 
We now give a synthetic description of when such spiral similarities exist in
nature.
10.1.3. Lemma. Let AB and CD be segments, and suppose X = AC ∩ BD
exists. If (ABX) and (CDX) intersect again at O, then O is the center of the
unique spiral similarity taking AB into CD.
Proof. This is actually just a matter of angle chasing. We have
]OAB = ]OXB = ]OXD = ]OCD
and similarly
]OBA = ]ODC.
That implies 4OAB ∼ 4OCD directly, which is sufficient. 
10.2. Miquel point of a cyclic quadrilateral. We begin with the following
surprising observation. Suppose O is the center of the spiral similarity sending AB
to CD as in the previous sections, meaning
4OAB ∼ 4OCD.
Then we have ]AOB = ]COD. This implies ]AOC = ]BOD. Moreover, we
have BO
AO CO
= DO ; so in fact we have
4OAC ∼ 4OBD.
In other words, spiral similarities come in pairs.
We can apply this observation to now get a famous theorem, which does not
refer to spiral similarity at all.
10.2.1. Theorem. Let ABCDEF be a complete quadrilateral. The four circles
(F AB), (F DC), (EAD), (EBC) concur at a point M .
Proof. Let M be the center of the spiral similarity sending AB to DC, hence
also sending AD to BC. Then Lemma 10.1.3 gives M lying on (F AB) ∩ (F DC)
as well as M lying on (EAD) ∩ (EBC), as needed. 
10.2.2. Definition. The point M is called the Miquel point of quadrilateral
ABCD.
In fact, we can get the following animated motion of the result.
10.2.3. Proposition. Let ABCD be a quadrilateral and let E = AB ∩ DC.
Consider variable points X and Y on lines AB and DC satisfying BX
AX
CY . Then
= DY
the circumcircle of 4EXY passes through the Miquel point M .
Proof. The spiral similarity mapping A to D and B to C should also map X
to Y , and we apply Lemma 10.1.3 again. 
 §10. COMPLETE QUADRILATERALS 34

10.3. The Miquel point of a cyclic quadrilateral. In the case where

ABCD is cyclic the point M has an explicit description.
10.3.1. Theorem. If ABCDEF is a cyclic complete quadrilateral with Miquel
point M , then M coincides with the inverse of AC ∩ BD with respect to the
circumcircle of A, B, C, D.
Proof. Let O denote the circumcenter. If we apply Proposition 10.2.3 with X
and Y as the midpoints of AB and CD we conclude that M lies on the circumcicrle
of 4EXY . This is the circle with diameter EO. Thus ∠EM O = 90◦ . Similarly
∠F M O = 90◦ . Thus M is the foot from O to line EF . Then we use Exercise 93. 
Exercises.
96. Exercise. Give a direct proof of Theorem 10.2.1 by angle chasing, without
referring to spiral similarity.
97. Exercise. Let ABCD be a quadrilateral, and let E and F be points on
sides AD and BC, respectively, such that EDAE
F C . Ray F E meets rays BA and
= BF
CD at S and T , respectively. Prove that the circumcircles of triangles SAE, SBF ,
T CF , and T DE pass through a common point.
98. Exercise (*). If ABCDEF is a complete quadrilateral with Miquel point
M , show that M is concyclic with the circumcenters of 4EAD, 4EBC, 4F AB,
4F CD.
99. Exercise (*). If ABCDEF is a complete quadrilateral with Miquel point
M , show that the feet of the altitudes from M to AB, BC, CD, DA are collinear.
100. Exercise. Let M be the Miquel point of cyclic complete quadrilateral
ABCDEF with circumcenter O. Show that the M is the second intersection of
circles (OAC) and (OBD).
101. Exercise. Let M be the Miquel point of cyclic complete quadrilateral
ABCDEF with circumcenter O. Show that M O bisects ∠AM C and ∠BM D.
 Index

areal coordinates, 21 power, 10

projective collineation, 30
barycentric coordinates, 21
radical center, 11
centroid, 5
cevians, 16 semiperimeter, 14
circumcenter, 5 signed area, 15
complete quadrilateral, 32 signed product, 10
complex conjugate, 18 signed quotient, 10
complex number, 18 similar, 8
concyclic, 7 Simson line, 9
cyclic quadrilateral, 7 spiral similarity, 32
symmedian, 31
directed angle, 5
directly similar, 8 unit circle, 19
displacement vector, 23

Euler line, 13
excenter, 7

generalized circle, 27
Gergonne point, 17

harmonic, 30
homogeneous barycentric coordinates, 22
homothety, 12

incenter, 5
intouch triangle, 5
inversion, 26
isogonal conjugate, 17
isotomic conjugate, 17

medial triangle, 5
Miquel point, 7, 33

Nagel point, 17
nine-point center, 13
nine-point circle, 13

oppositely similar, 8
orthic triangle, 5
orthocenter, 5
orthogonal, 28

polar, 30
pole, 30

35